IBM stock becomes the next victim of AI disruption

For years, the bull case on IBM (IBM) has sounded like this: as long as the world runs on COBOL, IBM gets paid.

That story took a punch this week.

Anthropic says its AI Claude model can automate COBOL modernization. Within hours, IBM’s stock dropped about $15 and briefly slipped under $230. Not because of some dramatic earnings miss, but because of what that claim implies.

COBOL still powers an estimated 95% of U.S. ATM transactions. It underpins hundreds of billions of lines of production code across banks, airlines, and government agencies. For decades, that legacy footprint has looked less like technical debt and more like an IBM tollbooth.

Now investors are squinting at a different possibility: what if COBOL modernization isn’t a long, expensive services cycle, but a software problem AI can meaningfully automate?

The drop was about control and the growing sense that AI might be taking some of it away.

The disruption behind the news: AI is turning consulting into software

COBOL has functioned like an annuity for IBM.

Legacy complexity has been its moat.

Claude just took a shot at that moat.

IBM’s consulting business thrives on systems that are too risky, too old, or too complicated for most firms to handle.

COBOL fits perfectly. It’s decades old, fragile, often poorly documented, and understood by a shrinking group of developers. That scarcity is billable. The fear of breaking a core banking system is billable. Multi-year modernization projects are very billable.

Anthropic is proposing a different model. Upload the codebase into Claude. Let the AI analyze it. Auto-generate documentation. Refactor step by step. Continuously test changes. Reduce the need for large human teams.

If that works even partially, IBM’s margins come under pressure.

If AI cuts modernization timelines by 30% to 50%, consulting budgets likely shrink. They don’t stay flat. If a bank planned to spend $200 million over five years to modernize COBOL systems, and AI reduces that to $120 million over three years, that $80 million difference doesn’t flow to IBM. It stays with the client or funds other digital projects.

IBM’s services teams often exist to pull through the real annuities: mainframe capacity, transaction-processing software, and long-lived support contracts that sit behind the COBOL estate. So the downside isn’t just “fewer billable hours.” It’s “fewer reasons to renew the stack.” If AI makes it meaningfully easier to leave (not just modernize), then even a small migration rate gets big fast: assume just 2% of IBM’s infrastructure-related run-rate gets displaced by faster off-mainframe rewrites, and you’re talking hundreds of millions per year in revenue at risk, the kind of compounding annuity math that markets punish instantly.

The bigger shift is who captures value. Power could move from traditional services firms to AI model providers and tooling platforms. IBM primarily sells expertise and consulting labor. Anthropic sells AI intelligence that can be reused at scale. One model scales with headcount. The other scales with computing power. Also Anthropic`s new AI tool recently disrupted software stocks.

That “95% of ATM transactions” statistic isn’t comforting. It signals vulnerability. The installed base of COBOL is huge and structurally similar across institutions. Large language models are especially good at spotting patterns in repetitive, rules-based code. Hundreds of billions of similar lines are ideal training terrain.

If Claude can reliably document, translate, and refactor even 10% of that installed base per year, the consulting industry built around manual modernization doesn’t disappear, but it likely gets repriced.

IBM won’t vanish. But its leverage changes. When clients believe modernization is no longer a 10-year lock-in, they negotiate harder and demand faster results.

What to watch next

Look for detailed case studies with real cost savings attached.

Watch how quickly CIOs run pilots and expand them.

Over the next 6 to 24 months, three signals matter most.

First, deployment speed. If a top 20 U.S. bank publicly says AI cut a core modernization timeline from 48 months to 24, that’s a major shift in expectations.

Second, pricing pressure. If IBM’s consulting growth slows while AI tooling revenue rises at companies like Anthropic, the market will continue to adjust valuations.

Third, regulation. Financial regulators require audit trails, validation, and explainability. If Anthropic can package those into enterprise-ready workflows, adoption speeds up. If not, incumbents keep their advantage longer.

This isn’t really about a single afternoon stock swing, but something bigger: does all that legacy complexity stay a steady cash machine, or does it just become fuel for the next wave of AI?

The real bet here is that the data wins. That COBOL modernization turns into the next big fight over who actually captures the value. The old-school services firms that’ve lived off this complexity for decades, or the new AI-native platforms built to automate it away. IBM (IBM) has a Disruption Score of 3.

Click here to learn how we calculate the Disruption Score.